UPTEC ES 18 039

Examensarbete 30 hpNovember 2018

Developing a Combinatorial Synthesis Database Tool

Luciano Quaglia Casal

Teknisk- naturvetenskaplig fakultet UTH-enheten Besöksadress: Ångströmlaboratoriet Lägerhyddsvägen 1 Hus 4, Plan 0 Postadress: Box 536 751 21 Uppsala Telefon: 018 – 471 30 03 Telefax: 018 – 471 30 00 Hemsida: http://www.teknat.uu.se/student

Abstract

Developing a Combinatorial Synthesis Database Tool

Luciano Quaglia Casal

Thin-film solar cell research is central to the electricity production of the near future.Photovoltaic technologies based on silicon have significant portion of the globalmarket and installed capacity. Thin-film solar cells are part of the emergingphotovoltaic technologies that are challenging silicon for a part of the electricityproduction based on solar power. These thin-film technologies, such as copperindium gallium selenide (CIGS) and cadmium telluride (CdTe), are lower cost andrequire less energy to produce, but also require rare materials. An alternative tothese technologies are thin-film solar cells based on more abundant materials. Todevelop these new materials at Uppsala University, combinatorial synthesis is used.This method produces a significant amount of data across different measurementmethods. The data needs to be analysed and combined to gather information aboutthe characteristics of the materials being developed. To facilitate the analysis andcombination of data, a database tool was created in MATLAB. The result is a programthat allows its User to combine energy-dispersive X-ray spectroscopy (EDS), Ramanspectroscopy and Photoluminescence spectroscopy measurements done on solar cellabsorber layers. Absorber layers are the section of solar cells where sun light isabsorbed, and electron-hole pairs are created. The program provides multiple figuresand graphs combining the different data collected, enabling the User to drawconclusions about the characteristics of the sample and its suitability as an absorberlayer. The combinatorial synthesis database tool created could be used forcombinatorial synthesis analysis of other material samples that are not necessarilyabsorber layers for thin-film solar cells. This report describes both the developmentof the tool and the code itself.

ISSN: 1650-8300, UPTEC ES 18 039Examinator: Petra JönssonÄmnesgranskare: Jes LarsenHandledare: Jonathan Scragg

ii

Executive Summary The combinatorial synthesis method is of great promise for the development of thin-film solar cells

based on abundant materials. Using abundant materials mitigates hindering factors for production and

development of thin-film solar cells such as copper indium gallium selenide (CIGS) and cadmium

telluride (CdTe). These factors are, for example, scarcity of materials and geographical problems such

as geopolitics. Therefore, the need for photovoltaic technologies based on abundant materials is clear.

The database tool created and programmed in MATLAB aims to facilitate and power the

combinatorial synthesis analysis used when researching thin-film solar cells built with abundant

materials. The program saves data in sample specific folders, analyses the data and presents the data

in figures and graphs. The data used is acquired from energy-dispersive X-ray spectroscopy (EDS),

Raman spectroscopy and Photoluminescence spectroscopy measurements. The sample type used

when developing the combinatorial synthesis database tool was CZTS, a thin-film technology with

an absorber layer based on copper zinc tin and selenide. The figures and graphs provided are mostly

centred on plotting Raman and Photoluminescence intensities versus material composition at the

measurement points. Most likely, in the future, there will arise new needs, thus the program is written

in a way that encourages its further code development.

iii

Populärvetenskaplig sammanfattning Global uppvärmning är ett välkänt fenomen och det akuta behovet av att minska utsläppen av

växthusgaser är ett faktum. Elproduktion från förbränning av fossilbränslen, med sitt

koldioxidutsläpp, är en stor bidragande faktor till växthusgasutsläppen. För att minska dessa utsläpp

kan alternativa energikällor och elproducerande tekniker användas. En mycket populär teknik för att

nyttja förnyelsebar energi är fotovoltaiska celler.

Fotovoltaiska solceller nyttjar solinstrålningens fotoner för att skapa en ström. Tekniken bygger på

att fotoner absorberas i absorptionsskiktet av solcellen. Med absorption menas att en foton exciterar

en elektron genom att överföra sin energi till elektronen. När denna exciteras hoppar den upp över

bandgapet, som är beroende på materialet som absorptionsskiktet består av, och lämnar ett hål.

Därmed har ett elektron-hål par skapats av solinstrålningen. Dessa elektron-hål par rör sig i motsatt

riktning genom solcellen och den yttre kretsen vilket ger en ström. I dagsläget är de flesta solceller i

användning baserade på kisel.

En konkurrent till kiselsolcellerna är tunnfilmssolceller som har en ökande popularitet.

Tunnfilmssolceller har flera fördelar gentemot solceller baserade på kisel. Precis som namnet

indikerar så är tunnfilmssolceller betydligt tunnare och kräver då naturligtvis mindre material än de

tjockare versionerna (exempelvis kiselsolceller). Dessutom är de mycket flexibla till sin natur, vilket

kan utnyttjas vid olika konstruktioner med varierande ändamål.

Problemet med dessa tunnfilmssolceller är att de tekniker som är mest prominenta just nu har vissa

materiella nackdelar. Koppar indium gallium selenid (CIGS) och kadmium tellurid (CdTE) är två av

de vanligaste tunnfilmssolcellsteknikerna i användning. Indium, gallium och tellurid är relativt

ovanliga material och är därför dyra och svåra att få tag på. Alternativet till detta vore att använda

ämnen som det finns rikligt av. Därför undersöks alternativa tunnfilmssolcellstekniker baserade på

vanligt förekommande material. Ett av dessa exempel är en teknik som använder ett absorptionsskikt

baserat på koppar zink tenn och selenid (CZTS). Denna teknik är dock i ett tidigt utvecklingstadie

och kräver mer forskning.

Forskning kring CZTS och andra material kan använda metoden kombinatorisk syntesanalys. Denna

metod bygger på att producera materialprover som har en varierande kemisk uppbyggnad baserad på

några olika ämnen. Sedan används olika mätningsmetoder för att samla data från den varierande

kemiska uppbyggnaden för att se hur den påverkar egenskaperna hos materialet.

Mätningar av materialproverna kan göras med energidispersiv röntgenspektroskopi (EDS),

Ramanspektroskopi och fotoluminiscens (PL). EDS används för att fastställa den kemiska

sammansättningen över materialprovet, medan Raman och PL används för att fastställa olika

egenskaper. Dessa egenskaper analyseras för att se vilken eller vilka kemiska sammansättningar som

är mest lämpade för att användas som absorptionsskikt i en tunnfilmssolcell.

Denna rapport beskriver utvecklingen av ett MATLAB program som har funktionen av en databas

för de data som samlas när EDS, Raman och PL experiment utförs på material producerade genom

kombinatorisk syntes, samt är ett analysverktyg. Programmet, vars namn är Combinatorial Synthesis

Database Tool (CSDT), kombinerar de olika experimentella data för att förenkla djupare analyser av

materialprovernas egenskaper. Resultatet blev ett program som producerar olika figurer och grafer

beroende på vad som önskas av användaren. Dessutom är programmet utformat på så vis att varje

delsteg beskrivs av programmet i löpande form för bättre pedagogik. CSDT ska användas i framtida

forskning av absorptionsskikt av Fastatillståndets elektronikavdelning på Uppsala universitet.

iv

Preface This project is the final work I will do as a student of Energy Systems Engineering at Uppsala

University and Sveriges Lantbruksuniversitet. With this my graduation will be complete and

therefore, I would like to give thanks to the people that have helped along the way. First of all, thank

you to my supervisor Jonathan Scragg that has been a great help and most importantly, challenging

to expand my reasoning regarding the project. Also, Jes Larsen, my subject reviewer, thank you for

providing invaluable input on the report and the project. Katharina Rudisch, thank you for always

being willing to lend a hand. My fellow master thesis students at the Solid-State Electronics solar cell

group Marika Gugole and Joakim Adolfsson. Thank you both for bringing joy and energy when the

project was most problematic and always being a source of laughter. Finally, my family Patricia, Juan

and Joaquin; thank you for providing constructive criticism from an outside perspective.

Luciano Quaglia Casal

Uppsala, October 2018

v

Abbreviations CdTe Cadmium telluride

CIGS Copper indium gallium selenide

CSDT Combinatorial synthesis database tool

CZTS Copper zinc tin selenide

EDS Energy-dispersive X-ray spectroscopy

GW Giga Watts

MW Molar weight

PL Photoluminescence

PV Photovoltaics

ROI Region of interest

SCR Space charge region

TCO Transparent conductive oxide

Table of Contents 1. Introduction & Theory ................................................................................................................ 1

1.1 Combinatorial Synthesis ........................................................................................................... 3

1.2 Energy-Dispersive X-Ray Spectroscopy ................................................................................... 4

1.3 Raman Spectroscopy ................................................................................................................. 5

1.4 Photoluminescence Spectroscopy ............................................................................................. 6

1.5 Master Thesis Scope.................................................................................................................. 7

2. Method ............................................................................................................................................ 8

2.1 Preparation ................................................................................................................................ 8

2.2 Code Structure Design .............................................................................................................. 8

2.2.1 Figures and Plots ................................................................................................................ 9

2.3 Implementation and Development ............................................................................................ 9

2.4 Data Management ................................................................................................................... 10

2.5 Evaluations .............................................................................................................................. 12

3. Results ........................................................................................................................................... 14

3.1 Combinatorial Synthesis Database Tool (CSDT) ................................................................... 14

3.1.1 First Section – Sample Information .................................................................................. 14

3.1.2 Second Section – Data Management ................................................................................ 15

3.1.3 Third Section - Analysis ................................................................................................... 16

3.1.4 Fourth Section – Results Presentation .............................................................................. 20

4. Discussion ..................................................................................................................................... 25

4.1 EDS fitting .............................................................................................................................. 25

4.2 Region of Interest Determination ............................................................................................ 25

5. Conclusions ................................................................................................................................... 28

5.1 Future Development ................................................................................................................ 28

5.2 New Insights ............................................................................................................................ 28

References ......................................................................................................................................... 30

Appendix 1 ........................................................................................................................................ 31

Appendix 2 ........................................................................................................................................ 37

Appendix 3 ........................................................................................................................................ 40

1

1. Introduction & Theory Long-term threats from greenhouse gas emission and specifically, carbon dioxide (CO2) from fossil

fuel combustion have naturally lead to an increased focus on renewable energy substitutions and an

increased awareness of environmental impact across the world. One of these renewable energy

sources is solar energy (Schmalensee, R., et. al. 2015, p. 1). The access to energy in useable forms is

closely related to local and global economic situations. Economic growth on a global scale means

that the demand for energy increases. Then the solution to global warming is not as simple as just

replacing existing demands for energy but meeting the increased need that accompanies this economic

growth. When considering the future electricity production, other aspects need to be considered, since

there are more than only economic constraints when planning long-term electricity and, for that

matter, energy production in general. Two major aspects are the political climate and the public

opinion. Nuclear power, despite being an emissions free energy source, is a clear example of this.

Nuclear power is a long-debated case and it shows how the renewable energy sources are a must in

the global energy production set-up of the immediate future.

Solar energy can be either thermal or photovoltaic (PV), although for the production of electricity PV

is the preferred technology. The basis of PV technology is that sun light is absorbed in a layer of

material with certain characteristics which creates electron-hole pairs. These electrons and holes

move in opposite directions through front and rear contacts, finally being reunited through an external

circuit. The most common PV technology is silicon-based cells which are around 0.2 mm in thickness.

Figure 1 shows a conceptual image of a Si solar cell. Sunlight enters the cell from the top, passing

through the anti-reflection coating and is then absorbed in the area labelled Space Charge Region

(SCR). The SCR is where the electron-hole pairs mentioned earlier are created and from where the

electrons move up to the front contacts and the holes move down to the back contact.

Figure 1: This figure shows a conceptual version of a silicon solar cell. The back contact, front contacts, anti-reflection

coating as well as the emitter and base layers are all shown. The space charge region is also shown with the arrows and

dotted lines.

Even though Si solar cells are the most common, thin-film solar cells are gaining prominence. Thin-

film solar cells such as cadmium telluride (CdTe) and copper indium gallium selenide (CIGS), which

have thicknesses of around 0.003 mm are among the PV technologies advancing on the market. A

couple of advantages of thin-film solar cells with respect to the common silicon devices are their

small material consumption and potential for flexibility. Si solar cells have the advantage of using

mostly abundant resources which is not the case for CdTe or CIGS. This partly explains why research

into thin-film solar cells with abundant materials is on-going and it is this research that this project

2

aims to facilitate. Specifically, research concerning the absorber layer, which is shown in figure 2 as

the blue layer.

Figure 2: This figure shows a conceptual version of a CIGS solar cell. The back contact, front contacts, transparent

conductive oxide layer as well as the substrate and absorber layer are all shown.

Thin film solar cells are considerably thinner than the regular silicon solar cells, and in fact, the

thickness of a thin film solar cell depends primarily on the substrate on which it is constructed. The

substrates are made of a sturdy transparent material such as glass or plastic. The transparent

conductive oxide (TCO) shown as green in figure 2 is a material that acts as the front contact. The

back contact, shown as grey, could be of the transition metal molybdenum. Sunlight passes through

the TCO window and is then absorbed in the space charge region created in the buffer (yellow) and

the appropriately named, absorber layer (blue). Then the electron-hole pairs move through the front

and back contact and create an electric flow through an external circuit.

The increased use of PV will be affected by materials and scaling issues when moving from small

scale to large scale production and installation (Schmalensee, R., et. al. 2015, p. 2). An estimate states

that 25 TW of zero-carbon energy will be needed, globally by the year 2050. In order to reach this

goal solar PV would need to scale up by one or two orders of magnitude (Schmalensee, R., et. al.

2015, p. 125). This scenario, and others, are provided by the International Energy Agency

(Schmalensee, R., et. al. 2015, p. 2). This coveted level of scaling would undoubtedly be hindered by

the availability of certain critical materials. Examples of these materials are indium, gallium, and

selenium for CIGS and tellurium for CdTe cells (Schmalensee, R., et. al. 2015, p. 125). These critical

materials are produced as minor by-products which means that the availability is dependent on the

main metals mined (Schmalensee, R., et. al. 2015, p. 126).

There are two primary ways of obtaining a higher output of these by-product materials. The first

option is improving the refining process to acquire a higher efficiency in the process. The second

option is to produce the sought-after material as a primary product (Schmalensee, R., et. al. 2015, p.

126). Either way, the alternatives require substantial investments, causing the final product cost to

simultaneously increase. So, the question of critical materials becomes one of trying to decrease the

necessity of the materials in question.

To tackle the necessity regarding materials with respect to energy security the focus turns to reducing

the amount used. That way, the production of electricity, for example, can be less affected by scarcity

and elevated costs. Either the material is substituted with another, more abundant material

(Schmalensee, R., et. al. 2015, p. 126) or the amount of material per peak watt is decreased. When

3

reducing the amount of material used there are, as expected, practical limitations (Schmalensee, R.,

et. al. 2015, p. 147). An example of a practical limitation is that, if an absorber layer is too thin, the

range of sunlight absorbed will decrease and, consequently, the efficiency of the cell also decreases.

Taking into consideration all the contributing factors described above, substitution of the critical

materials is worth exploring. This substitution should be done with materials of more abundance or

at least cheaper to acquire. Using abundant materials will counter-act the issues listed earlier, such as

the economic aspect and the limited resources. Another potential advantage of using abundant

materials could be mitigating geopolitical factors by reducing the chance of monopolization of

commodities. Toxicity is also a point to consider when choosing materials for thin-film solar cells,

such as the cadmium in the CdTe solar cells which is a toxic element. Two examples of PV

technologies that use abundant materials, instead of the critical ones stated earlier, are perovskite and

copper zinc tin sulphide (CZTS) (Schmalensee, R., et. al. 2015, p. 147).

Most likely, the future will require a mix of different PV technologies where both the existing ones

and those currently under development, or even unknown at the moment, are needed. There is also a

natural need for new solar cell production that results from technology increasing efficiency and life

cycle periods improving. Older existing modules in use can wear out or stand in the way for more

efficient and technologically advanced options. The production step is an important part of this

renewal and expanse of solar cell modules. This also needs to be made more efficient and different

solar cell technologies can aid these improvements. As explained earlier, there are many aspects that

dictate the need for materials with better properties for solar cells, but this is also true if new functions

or concepts become relevant. Regardless, there is always the need for more robust modules with

longer lifetimes to diminish the amount of material used. Central to the issues surrounding solar cells

are the absorber layers. The absorber layer technologies available and those in development can

positively affect economic and resource issues by switching focus from problematic materials to more

abundant and accessible substitutions. Therefore, the need for new absorber layer material

compositions will not diminish in the foreseeable future.

1.1 Combinatorial Synthesis The process of developing PV technologies is difficult for various reasons. The characteristics

required for the absorber layer are dependent on chemistry and physics that is sometimes hard to

predict with accuracy. For example, the absorber layer should have properties that minimize the

electron-hole recombination and maximize light absorption. These two characteristics allow for better

utilisation of the incoming photons from the sunlight. New materials can contain many different

elements. Defects are crucial to solar cells and they are highly sensitive to combination and

proportions of the elements involved. Defects influence the doping of materials and the recombination

rates which affects the efficiency of the solar cells. Controlling the defects is called defect engineering

and is essentially what finding an absorber layer requires. This means that a vast composition space

needs to be tested to find the optimum composition, if it even exists.

The way of testing materials was until fairly recently reliant on trial-and-error processes, which are

slow and serendipitous according to Hideomi Koinuma and Ichiro Takeuchi (2004, p. 426). The

pharmaceutical industry developed a new way of testing which was transformed to suit engineering

and material sciences to accelerate the procedure of studying new compositions. The method is called

the combinatorial approach and essentially means that many different compositions can be studied

simultaneously with a single experiment instead of one single experiment for each composition. Even

though a single experiment can be used, it will still be a set of repeated measurements at different X

4

and Y coordinates later correlated. The information collected is utilized to evaluate the way the

different properties are affected by the changing compositions.

The CZTS samples were created by sputtering with three different targets. These targets were

triangularly and evenly separated, and they contained CuS (copper and sulphur), ZnS (zinc and

sulphur) and SnS (tin and sulphur) respectively. The sputtering process deposited copper, zinc, tin

and sulphur on a glass substrate with back contact coating. To get a compositional gradient on the

samples, they were not rotated during the sputtering. The next steps were cutting the samples to a

smaller size and then annealing in sulphur rich atmosphere to crystalize the desired compound;

secondary compounds can also form during this process. Finally, a low temperature thermal treatment

was used to enhance the Cu-Zn ordering of the crystal structure. (Adolfsson, 2018)

Applied to thin-film research, the combinatorial approach is used when characterising combinatorial

libraries and composition spreads by spatially varying or selective deposition (Koinuma & Takeuchi

2004, p. 426). This means that many different elemental compositions are tested for characterization

during one experiment by varying elemental composition in one sample. This characterization or

property identification of samples can be done with different spectroscopies such as Raman and

photoluminescence in combination with energy-dispersive X-ray spectroscopy (or other composition

measurement), briefly described here below.

1.2 Energy-Dispersive X-Ray Spectroscopy The combinatorial approach can be accompanied by energy-dispersive X-ray spectroscopy (EDS).

This type of spectroscopy is useful for analysing and ascertaining the chemical composition of a

sample. It is based on having an electron-beam irradiating areas of the sample. These electrons can

transfer energy making other electrons leave the sample. Having an empty state leads to another

electron reducing its potential energy to drop down from an elevated state and occupy the empty

position. The energy emitted can have the form of X-ray quantum which is then detected and recorded

(Abou-Ras, et. al. 2011, p. 305). This concept is shown in figure 3 where the green arrow represents

the electron-beam ionizing an atom and the leaving electron is shown by the red dot. The empty space

represented by the hollow red circle is filled by an electron shown as a red dot which emits energy

expressed with the orange arrow. The image itself is of a theoretical atom in the sample. The different

elements have different characteristic wavelengths (or X-ray lines), which are used to identify the

elements present in the sample. The result of this evaluation will then be translated into the weight

percentage of the elements that make up the sample in question.

Figure 3: A conceptual image of the EDS technique. Ionization of atoms leaves empty spaces that are filled by electrons

emitting energy that is then detected. The red dots are electrons, the green arrow represents the energy from the electron-

beam, the purple arrows represent the path of electrons and the orange arrow is the emitted energy. The blue dot is the

nucleus and the blue circles are the energy levels or shells.

5

1.3 Raman Spectroscopy After having specified the composition of a sample with respect to the elements the next step is to

analyse the structure and crystallization, and which compounds are present besides the intended one.

To gather this type of information Raman spectroscopy can be used. The technique is one where

inelastic scattering of photons is measured to observe vibrational excitations of the materials present.

From the Raman spectrum different aspects of the lattice can be discerned. Depending on the position

and presence of peaks, different phases can be identified. They could be main phases or secondary

phases of the sample. This in turn gives information of how the material responded to its production

process. (Abou-Ras, et. al. 2011, p. 365)

By using lasers with different wavelengths (infra-red, ultra violet, etc), the information acquired in

the form of intensity and Raman shift, which is the difference between excitation energy and detected

energy, can be interpreted more easily. Different lasers are used to obtain “resonant” conditions,

where the laser energy closely matches the band gap of a given phase present in the material, which

results in a much stronger signal.

For this project, the data collected from these Raman measurements are coordinates, wavenumber

and intensity. The wavenumber is the Raman shift and the inverse of wavelength. Figure 4 below

shows how a certain point on a sample will have varying spectra depending on the wavelength of the

laser.

Figure 4: An example of the output from Raman spectroscopy. Three lasers with wavelengths 785, 532 and 325 nm

respectively gave the output above. Intensity is given on the Y-axis and Raman shift is given on the X-axis. These

measurements were all done on sample 12798A at the coordinates X = 44 mm and Y = 25.25 mm. The sample is of the

CZTS type.

In Raman, the resonance effect when the laser energy is close to the band gap energy of the material

being studied causes a dramatic signal enhancement. We have several different laser energies and the

various secondary compounds in the samples which have different band gaps. Thus, resonance can

be achieved for several compounds, for example for CZTS and CTS [band gap 1.5 to 1.6 eV] with

the 785 nm laser and ZnS [3.6 eV] with the 325 nm laser. The difference between resonance and non-

resonance can be seen in figure 4 with the there being clear differences between the three curves.

6

CZTS samples can be analysed with secondary order parameters from the Raman spectra that show

the copper zinc disorder. The secondary order parameters are ratios of Raman intensity at specific

Raman shifts. (Davydova, A., et. al. 2011, p. 10) The ratios show trends of copper and zinc ordering.

The sample properties change due to the slight structural changes caused by the copper and zinc

disorder. This is what leads to changes in the peak heights and this fact is used to acquire the

secondary order parameters used for disorder analysis.

The secondary order parameters are defined by equations 1 and 2. The first equation is Q1 which is

the ratio between the Raman intensity of CZTS main peaks two and three. The main peaks of CZTS

are expected Raman intensity peaks for this type of sample. Q2 is described by equation 2 and is the

ration of Raman intensity of CZTS main peak one and the sum of the Raman intensity of CZTS main

peaks three and four. All main peaks are specified in table 1 later on, as well as the intervals that

correspond to them.

𝑄1 = 𝑀𝑎𝑖𝑛 𝑃𝑒𝑎𝑘 2

𝑀𝑎𝑖𝑛 𝑃𝑒𝑎𝑘 5 (1)

𝑄2 = 𝑀𝑎𝑖𝑛 𝑃𝑒𝑎𝑘 1

𝑀𝑎𝑖𝑛 𝑃𝑒𝑎𝑘 3+𝑀𝑎𝑖𝑛 𝑃𝑒𝑎𝑘 4 (2)

1.4 Photoluminescence Spectroscopy Photoluminescence (PL) spectroscopy is based on the emission of light, following absorption of

photons. In the case of the PL measurements that provided the data used in this project a

monochromatic light source was used. The light, typically from a laser, excites electron-hole pairs

leading to electrons reaching higher energy levels in the material. Light is then re-emitted when the

electrons fall from higher states into lower unoccupied ones; and it is this emission that is measured

(Abou-Ras, et. al. 2011, p. 151). Figure 5 shows how this process happens. The electrons and the

holes they leave are red dots and red circles respectively with the orange arrow showing the emitted

energy detected. What the photoluminescence measurements provide is a pattern of electron-hole

recombination for a sample. The information gathered from this leads to conclusions about the opto-

electrical properties of the material measured. Essentially, the stronger the intensity of the measured

radiative recombination, the better the material quality is, because this indicates that other types of

(non-radiative) recombination, which are harmful to solar cell operation, are less significant. Data

from these measurements are coordinates of the measurement points, intensity of radiative emission

and the photon energy.

7

Figure 5: The figure shows how incident energy (green arrow) excites an electron which reaches a higher energy level. For

example, from the valence band (black line) to the conduction band (blue line). The purple lines show the electron (red dots)

paths with the red circles representing the previous position. When the electrons go down to a lower energy level energy is

emitted (orange arrow) and detected. This process is shows the photoluminescence spectroscopy concept.

In summation, the samples are characterized by using EDS to specify chemical compositions, Raman

to identify phase structure and quality of crystallization, and PL to observe opto-electrical properties.

These results give feedback for optimization of the material synthesis. Of course, other spatially

resolved measurement techniques could be implemented to obtain further information and additional

measurements such as SEM characterisation to examine grain structure could be of interest. However,

this is outside the scope of this report. By the process of using PL, Raman and EDS one can ultimately

proceed to analysing if a material is worth researching further with regards to feasibility as an absorber

layer for thin-film solar cells.

1.5 Master Thesis Scope The aim of this master thesis was to design and develop a database tool written in MATLAB that

stores and analyses the data produced from measurements characterizing samples of compounds

produced with the combinatorial approach. Since an actual database was not used, the database

function was performed by a set of folders. The database tool will be referred to as the Combinatorial

Synthesis Database Tool (CSDT). The goal was to facilitate combinatorial synthesis analysis with

regards to the large datasets with hundreds or even thousands of data points or spectra created by the

measurements. To expedite the analysis process with automation was central to reach the goal and

aim of the project. In other words, the combinatorial synthesis database tool needs to take large

datasets and combine them to produce useable results in the form of figures and graphs.

The scope of the project and subsequent report was focused on a product as the end goal as opposed

to only an investigation. The scope includes creating a program that can handle basic functions of a

database, format and analyse data acquired by experimentalist at the Solid-State Electronics division

at Uppsala University, and finally produce documentation and figures to accompany the analysis.

Making samples was outside the scope of the project. Nor did it include taking measurements or

performing experiments on these samples. Those activities were performed primarily by Jonathan

Scragg, Katharina Rudisch and Joakim Adolfsson at the Solid-State Electronics division.

8

2. Method The process of the project was modulated into four steps or parts, where the last step was producing

this document. Besides the four steps originally intended (project preparation, code planning,

implementation/development and the report) two other activities were added as part of the last step.

The first being data management depending on data file formatting. The second activity was to choose

optimal parameters for the analysis of data.

2.1 Preparation The first step of the project was a literature study of different measurement methods performed on

the absorber layer samples. Also, the reasons for using PV, as described in the background section

above, was once again stated. Finally, the actual aim of the project and desired result was specified.

2.2 Code Structure Design The code structure design consisted mainly of discussions with the project supervisor to identify the

desirable functions of the code. Hence, the code was mapped out on paper as a way of coming to

terms with the coherence and content of the different aspects and parts of the program. The key benefit

of this step was to minimize the risk of having to make significant changes midway through the coding

process due to missed or unforeseen connections between modules or sections of the code.

The following list shows the aspects and functions that were desired for the CSDT. Naturally, this list

could be subject to changes due to emerging needs and so forth. The nature of this flexibility could

have become a problem of delimitation. Therefore, priority had to be established and things classed

as low priority where to be done if time was available.

1. Basic database functions

a. Read information about one or more samples

b. Add new sample

c. Update existing sample by adding or removing data files

d. Delete a sample

2. Preparing data for analysis

a. Read in data files

b. Format data to bridge the gap between measurement standards and MATLAB

calculations

3. Analysing data

a. Calculations with data

i. Fit EDS data to acquire polynomial fitting curve

ii. Integrate around main peaks and other phases for Raman

iii. Integrate PL curves

iv. Find composition at Raman and PL measurement points

4. Result presentation

a. Figures and plots

i. Comparing EDS data and Fitted EDS data

ii. Raman peak intensities vs. composition

iii. Raman peak intensities vs. X and Y coordinates

iv. PL intensity vs. X and Y coordinates

v. Secondary order parameters vs. composition

vi. Secondary order parameters vs. X and Y coordinates

vii. Raman spectrum for specific measurement points.

9

The basic database functions were to be provided in the MATLAB interface. Therefore, the ability to

change/rewrite and save files/folders was to be provided by the CSDT. However, this step itself could

be handled by the User navigating folders normally as a back-up measure. Priority-wise this was

therefore not as high priority as data preparation, analysis and presentation.

2.2.1 Figures and Plots

When characterizing and investigating combinatorial synthesis samples, different types of plots and

figures are required. The aim is to find patterns or consistent behaviours at different areas of the

samples, and tailored plots are a powerful tool in this regard. Combining the results of the above-

mentioned EDS, Raman and PL measurements is advantageous. Especially considering that the

alternative closest to hand is to look through each individual spectrum pertaining to each measurement

point. By associating observed properties from Raman or PL with the material composition, the

researcher can start to understand the underlying physical processes accountable for the different

properties. Therefore, three-dimensional plots with combinations of either Raman or PL and EDS

were sought after as a way of finding those patterns and properties.

2.3 Implementation and Development The database characteristics of the code arise from the fact that data is stored in a folder system with

the aim of minimizing the interaction between the User and the actual folders. When developing the

code, the previous statement was fundamental. Figure 6 shows a schematic image of the directories

and hierarchy of the database system.

Figure 6: A schematic overview of the database hierarchy. Each level is contained within the level above with the Main

Database Folder being the parent (top) folder containing everything else.

To reach the goal of minimal interaction between the actual folders and the User the main script had

to allow the User to modify, add and delete single files and even entire samples through the code

interface. Thus, this requirement was specified as: the User shall only need to use the command

window and menu interaction while running the main script in MATLAB. This had to be enough for

the User to acquire all the information, figures and analysis sought. An important particularity of the

CSDT was that the intended target group of Users was small and therefore involved in the

development. Realizing that the demand of functions, analysis and figures could change during, and

after, the course of the project a certain flexibility was necessary.

10

2.4 Data Management What EDS, Raman and PL have in common is that the data provided is always presented at specific

points on the sample, and the various measurements need to be compared at the same locations. When

evaluating Raman and PL data, this is not considered a problem because the measurement points can

be specified precisely by the experimentalist. For EDS, on the other hand, the measurement points

are automatically generated by the measurement software and cannot always be made to coincide

with the points measured with the other techniques. To tackle this issue the EDS data had to be fitted

as a continuous surface for each element of the composition that was of interest to the samples

handled. After that, the composition of any given point could be interpolated.

Figure 7 below shows the basic layout of a sample with its own coordinate system. The origin (red)

is chosen to be a specific corner of the sample. However, any measurement starting point is shifted

inwards due to the damage the edges have, which is caused by the cutting a sample goes through. The

green point represents this starting point and the purple one represents the last measurement point.

The black points represent the rest of the hypothetical measurement points. The coordinates (X and

Y) for a sample are measured in millimetres, while it is in micrometres for Raman and PL

measurements.

Figure 7: A schematic view of a sample set up for measurements. The origin point (red) is set on the actual top left corner

of the sample. The starting point (green) of the measurement is moved in from the origin. The last point (purple) is

somewhere on the bottom right side of the sample. The rest of the measurement points (black dots) have a frame between

the outer ones and the edges of the sample. The coordinate axes are orientated with X to the right and Y downwards which

is the standard for the sample coordinate system.

To fit the EDS data the weight percentages are first normalized to only include the wanted elements.

One example of this was the CZTS type of sample which had calcium (Ca), copper (Cu), zinc (Zn),

tin (Sn) as well as the total of the weight percentages. The issue with this was that the Ca was not to

be used and was subsequently removed. Because of this removal, the total no longer equalled 100 %

and an additional normalization was required. The next step was to convert weight percentage to

atomic percentage and once again normalize to avoid roundoff errors introduced during this

conversion. With the data formatted for future use, new EDS files were created with only atomic

percentages for each relevant element.

To convert from weight percentages to atomic percentages equation 3 bellow was used.

𝑎𝑡%𝑎 = 100 ⋅ (𝑤𝑡%𝑎

𝑀𝑊𝑎)/(

𝑤𝑡%𝑎

𝑀𝑊𝑎+

𝑤𝑡%𝑏

𝑀𝑊𝑏+

𝑤𝑡%𝑐

𝑀𝑊𝑐) (3)

The subscripts a, b and c represent the different elements. MW is the molar weight, wt% is the weight

percentage and at% is the atomic percentage.

11

One aspect of the EDS data that was simpler than its Raman and PL counterparts was the fact that

each measurement point only had one value for each column. Figure 8 shows an excerpt of a EDS

data file before the formatting has taken place. In the figure each row corresponds to a measurement

point on the sample. In other words, the only information not provided by the data file was how to

orientate the data according to the sample coordinate system. Naturally, there would still only be need

for one row of values per measurement point for the formatted file. Still, this coordinate information

would need to be added by a User later.

Figure 8: An example of how an EDS data file looks before it is formatted. This example is of a CZTS sample. It has five

columns with values corresponding to the weight percentage for Ca, Cu, Zn, Tn and the total. The number of rows with

values will be equal to the number of measurement points of the EDS experiment.

When performing EDS experiments the orientation of the coordinates on the sample follow figure 9.

The origin (red point) is set in the top left corner of the sample with the starting point (green point)

of the actual measurement shifted in. The yellow section represents the measurement area and the

purple point is the last measurement point. The reason for direction of the axes is due to the

instrument, and the distance between measurement points is measured in millimetres.

Figure 9: A schematic view of a sample set up for measurements. The origin point (red) is set on the actual top left corner

of the sample. The starting point (green) of the measurement is moved in from the origin. The last point (purple) is

somewhere on the bottom right side of the sample. The measurement area (yellow) has a frame between itself and the edges

of the sample. The coordinate axes are orientated with X to the left and Y downwards.

As can be seen from the figure 9 the measurement needs to move to the right and downwards to reach

the last point. This means that the coordinates in X will decrease while the coordinates in Y increase.

Therefore, the coordinates for the origin, starting point and the last point needed to be recalculated so

the origin would be at (0,0). X direction points to the right instead while the Y direction stays the

12

same after recalculation. This specific way of defining the coordinate system is used to make it

compatible with the Raman and PL measurements.

The Raman data provided was given in four columns (see figure 10). The first two corresponded to

the X and Y coordinates in micro meters. Wavenumber and intensity were the third and fourth column

respectively. When applying these data, the issue arose due to the fact that a whole spectral range was

saved continuously for each point of measurement. This meant that each single row was no longer

the only row corresponding to a single specific measurement point. Instead, a collection of rows

together, corresponded to a single measurement point. Therefore, when importing these data to

MATLAB, cell arrays were needed. Each cell element corresponded to a different measurement point

and each cell contained the whole spectral range of the measurement. This method was also used for

PL data import. When placing the matrices or tables into the cell elements the order needed to match

the measurement order. For example, if all the points in the X direction were measured before moving

in Y direction the process of importing into cells needed to follow that same order. Figure 10 shows

that the first coordinate that changes is in the X direction and subsequently all points in the X direction

were measured before moving to the next level in the Y direction.

Figure 10: An excerpt of a Raman data file (sample 12819A in this case). The values in columns 1 and 2 are coordinates

given in micro meters. They represent the position of the measurement point. Columns 3 and 4 are the wavenumber and

intensity measured. The red highlight shows the first time the coordinates shift to a new measurement point. The first

direction that changes is the X direction.

2.5 Evaluations The evaluation of EDS data had to be based around an understanding of the chemical composition at

points of interest. Therefore, a curve fit was applied for each element, with the aim to interpolate

compositions between measurement points. Using a curve fit will also automatically lead to noise and

outlier points to be filtered out. To accomplish these tasks the curve fitting tool provided by MATLAB

was used. More specifically, the polynomial method was used for two variables. To decide the orders

of the polynomial the surfaces corresponding to the different orders had to be visually compared to

13

the actual values. The aim was to, at a later part of the code, use the polynomial expressions with the

coordinates matching the Raman and/or PL. This way the User could match sample characteristics

with chemical composition. Acquiring these coordinates was quite straight forward due to them being

provided in the data.

When evaluating the peaks within the Raman spectra, peak intensity was considered. Integration of

the peaks was used to compare the different regions and measurement points of the sample.

Characteristic Raman peaks are found at certain wavenumbers depending on the sample type and

composition. Since the measurement calibration is not made for every data point, a small shift of

these peaks can occur, the width of the peaks can vary as well. Therefore, a region of interest (ROI)

needed to be decided. Deciding the ROI required visually checking different spectra of a sample and

looking for appropriate intervals around the peaks. These intervals need to encompass the peak

without incorporating other nearby peaks. Due to the nature of the code and user interaction, a region

of interest had to be specified for all peaks individually. The PL on the other hand only required an

integration of the whole spectral range for each measurement point.

Other evaluations were performed in a more continuous fashion without any extensive analysis. These

were by nature less rigorous than the other two explained previously, such as the choice of functions

and code layout. These choices could have incidence on the run-time of the scripts or the ease of

access to certain variables or information stored in the MATLAB work space. Seeing as practicality

was central to the process of programming the scripts, the less critical design choices have not been

documented.

14

3. Results The program architecture was made up of a main script, specific dedicated scripts and function files.

Function files are used by main program providing certain information in return getting calculation

results and specified parameters. The specific dedicated scripts are run from the main script when

needed. All these code files are shown as an appendix.

3.1 Combinatorial Synthesis Database Tool (CSDT) The program, as well as its main script, has a modular design with limited choices for the User. The

reason for this design is to comply with the user-friendliness required while still enabling the

acquisition of all data and input needed. The four sections of the main script are placed in the

following order: Sample Information, Data Management, Analysis, and finally, Figures, Graphs and

Results. The navigation between these sections was set to be restricted by a sort of chronology where

needed. An obvious case being plotting figures without analysed data. If there were no calculations

done there would be no data to plot. The run order of the program follows the modular section layout

shown by figure 11.

Figure 11: The boxes represent the sections that make up the Main Database Script with the run order shown by the red

arrow. The blue arrows pointing towards the boxes describe the input needed from the User. Blue arrows pointing upwards

from the boxes show the output provided by the program.

When the User already knows the details of the samples that will be used, the first and second section

would most probably be unhelpful and unnecessary. Therefore, the option of skipping those sections

is provided.

3.1.1 First Section – Sample Information

The first section is an interaction between the User and the program where all the relevant information

describing a given sample is provided. When the User chooses between a complete list of samples or

a list of a specific type of sample, a table is provided with identification, date added, additional

comments and the kind of measurements included in the sample folder (EDS, Raman and/or PL data

files). A comments field allows the User to provide additional information to easier identify the

sample of interest. The complete list is just a compilation of the specific sample information files and

therefore provides a table with information about all samples included in the database sorted by the

date the sample was added, see figure 12. This is meant to give indications if more data needs to be

added to the database. Later in the program the User will be able to see specifically what options are

available for a certain sample, but at this early stage the information is meant to save the User time.

15

Figure 12: Example of the complete list of sample information. Column 1 shows the sample type. Column 2 shows the

sample identification that consists of five digits and a letter. Column 3 gives the date the sample was added to the database.

Column 4 displays a comment provided by the User for that specific sample. The last three columns (5, 6 and 8) show

whether EDS, Raman or PL files are saved in the sample folder (either Yes or No).

All sections have a navigation menu where the User is asked if there is a desire to move on to the next

section. This is designed to give the User the option of trying something again if, for example, a

mistake was made. Menus also have the option of skipping a step without any action, which is useful

since sections one and two are run automatically. Thus, if the User has no need for the Sample

Information section or the Data Management section he or she can move on by clicking “none” or

“nothing” in the appropriate menu.

3.1.2 Second Section – Data Management

The second section provides a menu where the User chooses between the options: create new sample

folder, update/edit existing sample data set, delete existing data set or sample, and nothing. Here

nothing means, as explained earlier, just proceeding to the next section.

The option create new sample folder is used to add a folder for a new sample and gives the User the

option of adding a comment for the sample list as shown in column four of figure 12. This is the

extent of the first option meaning that, to actually add the data, another option needs to be chosen

later. There is a fail-safe programmed that lets the User know if the sample already exists or if the

sample name was left empty. It is worth mentioning that, since the CSDT is expected to be used by a

limited number of researchers with good knowledge of the available data and ongoing activities, the

user/code interactions are not infallible. In other words, mistakes can be made, but are not expected

from the User. Also, only a few faults are handled automatically by the code. If the name of the new

sample folder is wrong the User will need to choose the option of delete existing data set or sample

and follow the instructions or manually enter the folder system and remove the folder. This second

option would also require the User to remove the sample information corresponding to that erroneous

sample both in the complete sample list and in the sample-specific list.

After adding a new sample, the next logical step would be to add the corresponding data sets.

Therefore, the User should enter the section of update/edit existing sample data set where the text

files with data will be moved to the correct folder. To accomplish this the User needs to place all the

data files in the main folder. It is the same folder where the Main Database Script is found. Then the

program will ask for a few copy and paste type inputs which will move these files to the correct folder.

Afterwards the User is given the option of amending the comment section described earlier. The User

is also prompted to change the EDS, Raman and PL columns shown in figure 12 from No to Yes if

necessary. This is where the naming of files shows its importance for the first time. The Users need

to name the files in an appropriate way to facilitate the quick realization of what is contained within

them. The list below gives a few suggestions concerning the names of the data files containing EDS,

Raman or PL measurements.

16

• EDS files should contain EDS and the sample name

• Raman files should contain the wavelength of the laser in [nm] and the sample name

• PL files should contain the same as Raman files but also PL

The final section of the Data Management step is update/edit existing sample data set where the

options are also quite self-explanatory. The User chooses between deleting a whole sample or specific

files. If the first option is selected the whole folder pertaining to a sample will be permanently deleted.

The lists will be updated by the comment being changed to include *DELETED* in the beginning as

well as turning the indicators of EDS, Raman and PL files automatically to No. If option two is

selected the User needs to choose the file or files to be removed from a list. Then the comments can

be amended, and the indicators of EDS, Raman and PL files changed if needed. This concludes the

database functions of the program. In summary, lists of sample information are provided so the User

can see if data is available for analysis and then the option of managing these data are given.

3.1.3 Third Section - Analysis

3.1.3.1 EDS Analysis

The third section is the analysis part of the code and requires the User to specify the sample that will

be analysed. In the usual case, the EDS analysis should always be performed first due to the nature

of combinatorial analysis that the programme is aimed for. Having an EDS analysis allows for

matching characteristics with composition, although this is not mandatory for the program to

compute. The EDS analysis askes the User to choose a file from a list corresponding to the sample.

If this file is to be used for the first time an import script will run and format this file and save it as a

new version with the ending “_Formatted” on the file name. The User will then be able to use the

formatted version hence forth while the original version is kept. The import script recalculates the

weight percentages of the elements of the sample to atomic percentages. It also removes the columns

that are unnecessary. Figure 8 showed how the EDS file hade five columns, the result of the

formatting is showed in figure 13 below. The three remaining columns are the atomic percentages of

Cu, Zn and Sn for the CZTS sample 12819A.

Figure 13: Fragment of a formatted EDS data file for CZTS sample 12819A. Columns one through three are the atomic

percentages of Cu, Zn and Sn respectively. Each row corresponds to one measurement point on the sample.

17

After the formatted file is chosen the data is imported and the fitting script

for EDS is executed. Next is the input of coordinate information of the

specific sample. Figure 14 shows the coordinate input menu provided by

the program. The first two lines are for the number of measurement points

in the X and Y directions. Lines two and three are for the origin

coordinates for X and Y respectively. All the coordinate points are to be

given in millimetres. The origin can be chosen to be a corner of the sample

as shown with figure 9. Lines five and six correspond to the first

measurement point which is offset with respect to the origin so as to move

away from the edge where the sample has damage from the cutting

process. Coordinates for the last point of measurement is to be entered in

the last two lines. The program calculates the distance between

measurement points as well as the coordinates for X and Y in the form of

two separate vectors for later use in the program. Figure 15 shows lines

61 to 65 from the EDSFit.m file and the result from this for-loop is that

three matrices are created. These matrices contain the atomic percentages

described before but they have been placed in such a way that they have

the same location as their respective measurement points as well as being

separated by element.

Figure 15: Lines 61 to 65 from the EDSFit.m file that show the placement of atomic percentages for their respective element

and in the sample’s layout. This means that the matrix is the same order as the sample. At_pc is the variable containing all

atomic percentages. Cu, Zn and Sn are the new, element specific atomic percentage matrices. nyEDS and nxEDS are the

number of points in Y and X for the EDS data file.

Once the element matrices are in place, the fitting process can begin. Choosing the degrees of the

polynomial fit required observing the surface created and the measurement points to evaluate the level

of similarity. Having a higher polynomial degree was not observed to be a concern with respect to

computational time, but carefulness is required to not overfit the data points.

Figures 16, 17 and 18 show surface fits from the EDSFit.m file for CZTS sample 12798A. Figure 16

is for Cu and has a polynomial fit with order four for both X and Y in the left image and order two in

the right image. Figures 17 and 18 are done with polynomial fits of order three in X and Y. They

correspond to the elements Zn and Sn respectively. The blue dots are the experimental values from

the EDS data file. Where the surfaces have areas without blue dots the measurement points are below

the fit. The X and Y axes are the X and Y coordinates of the sample in millimetres. The Z axes are

atomic percentages of the element plotted. When looking at the surface and measurement plots it

becomes clear that for CZTS samples the Cu was more demanding than Zn and Sn. The fact that Cu

was more difficult to fit is further illustrated by the difference between the plots in figure 16a and

16b. Also, worth noting is that the root mean square error for the fits were all observed to decrease

with a higher order of polynomial fit but that it always remained relatively large due to the inherent

scatter in the EDS data.

The scatter and uncertainty of the empirical values can be explained to be a result of the production

of the samples. There is a rearrangement of the crystalline structure caused by the annealing step.

Without this rearrangement a smooth gradient would be expected. The uncertainty is implicitly

Figure 14: Pop up menu where

coordinate information is entered by

the User and applied to the analysis

of the current sample.

18

handled by the polynomial expressions that lessen the effects of single outliers. There is also an

uncertainty when performing the EDS measurements that cannot be avoided, such as statistical errors.

Figure 16 a & b: The surface fit of Cu atomic percentage compared to the experimental values of atomic percentage for

CZTS sample 12798A. The X and Y axes are the sample coordinates at X and Y. The polynomial expression of the surface

has an order of four in both X and Y for the left image (a). For the right image (b) the order is two in both X and Y.

Figure 17: The surface fit of Zn atomic percentage compared to the experimental values of atomic percentage for CZTS

sample 12798A. The X and Y axes are the sample coordinates at X and Y. The polynomial expression of the surface has an

order of three in both X and Y.

19

Figure 18: The surface fit of Sn atomic percentage compared to the experimental values of atomic percentage for CZTS

sample 12798A. The X and Y axes are the sample coordinates at X and Y. The polynomial expression of the surface has an

order of three in both X and Y.

When another sample type is incorporated into the program the polynomial degrees for that sample

type need to be decided. In practice, it means that the User needs to try different orders of polynomial

fits for each element contained in the sample that needs to be fitted. Then the User can decide what

is optimal for that sample type. Among the samples evaluated, there is no great harm in using order

four for all surface fits. What happens is that the computation will take slightly longer. The parameters

that are unnecessary when using higher order polynomials become approximated to zero with

coefficients close to zero without a serious impact on the fit.

3.1.3.2 PL Analysis

The PL part of the analysis is quite simple, and the only aspect worth mentioning is the fact that the

data files are large and therefore, are usually split into multiple files. The ImpPL.m file is executed

to import the PL data and manage the multiple files. The program first asks for the number of files

corresponding to the data set and then asks for the specific files with command line prompts. The

analysis itself is just the sum of the intensities for the measurement point. In other words, the

integration of the intensity curve. However, that the data is split, and subsequently needs to be

compiled into a new matrix with the whole data set in a consecutive manner. The intensities are placed

into a cell array with the cells in the correct place with respect to the sample orientation (see Appendix

3, ImpPL.m, rows 29 to 39). After the cell array is made, the sum operation is carried out. Finally,

the actual coordinates of the measurement points are extracted and placed into matrices for X and Y

respectively. These matrices have two dimensions with either the columns or the rows being repeated

to match the number of measurement points. The coordinates are then used to find the element

composition at the PL measurement points. This is done by using the fit polynomial with the X and

Y coordinates as input. These results can be used to acquire different figures in the last section of the

program.

3.1.3.3 Raman Analysis

The last analysis option is for Raman and the section starts with the User choosing a data file and

specifying the wavelength of the laser in nanometres. The data is then split up into cells in a cell array

that is orientated in the same way as the sample. This follows the same principle and method as the

corresponding section of the PL analysis. To split up the corresponding spectral data, the program

identifies the number of points in X and Y, as well as the number of rows. Dividing the number of

rows by the product of number of points in X and Y gives the number of data rows per measurement

point (see Appendix 3, dbtool.m, rows 333 to 346). At this stage the program also takes into

20

consideration that the measurement points for Raman and EDS do not coincide necessarily. The

coordinates for Raman are extracted and applied to the fit-polynomial provided by the EDS analysis

so the compositions are specified at the points of interest for Raman.

Certain sample types, and CZTS specifically, need composition ratios, as explained earlier, for the

figures and graphs. These ratios are specified in function files and require an input of atomic

percentages for the elements involved. After using the polynomial expression to calculate the

composition at the measurement points of Raman (see Appendix 3, dbtool.m, rows 3065 to 366) and

PL (see Appendix 3, dbtool.m, rows 302 to 304), these values are used by the CZTS function file to

calculate the X axis ratio and the Y axis ratio that will be used in the final section of the code. These

function files also need to specify the position of the Raman peaks of the main phases and the

secondary phases. The ratios for the secondary order parameters are also specified in the function file

but not used. Their only use is instructional. In the Raman calculation script, however, the ratios need

to be provided and will be used by the program. The Raman calculation script takes each cell

containing the intensity and integrates around the positions of the main peaks and secondary phases

within a ROI (region of interest).

The ROI is an interval which is meant to encompass the peak that will be integrated. The first option

was having a set ROI for all peaks and phases. Arbitrarily, seven wavenumber points in each direction

of the estimated maximum position, was chosen. In other words, if the maximum is expected to be

found at 370 cm-1, the ROI is from 363 to 377 cm-1 for that specific peak. Having a set ROI could be

disadvantageous for various reason. One reason being that peaks and phases can coincide within the

ROI, meaning that the intensity is erroneously accounted for more than once. A second reason why

having a predefined ROI can be problematic, is that the whole peak might not be encompassed by the

interval. Therefore, an investigation into a more reliable set of ROI intervals was performed.

There were four CZTS samples available for use when deciding the ROI. Of these four samples all

had Raman data for the 785 nm and 325 nm laser measurements. Only one sample hade Raman data

for the 532 nm laser. As explained earlier, different wavelengths show different peaks and phases.

That meant that phases only shown by 532 nm lasers have a ROI decided from only one set of data.

The ones for 785 nm and 325 nm on the other hand, are decided from combining all four different

data sets. Table 1 shows the ROI boundaries around each main peak and secondary phase

wavenumber. To measure the main CZTS peaks one through five and secondary phase Cu3SnS4 the

785 nm laser is used which means that the boundaries for these wavenumbers cannot overlap. The

laser with 325 nm shows ZnS and CuS while laser with 532 nm shows CuS, SnS and SnS2. All the

phases shown with lasers 532 and 325 are secondary phases.

Table 1: The table shows the lower and upper boundaries for the CZTS region of interest. P1, P2, P3, P4 and P5 are the

main peaks and the wavenumber is shown below the peak name. The five latter columns are the secondary phases that are

expected for the CZTS sample type. The values shown are wavenumbers with unit cm-1.

3.1.4 Fourth Section – Results Presentation

The fourth and final section of the main script encompasses the generation of the figures and graphs

section that facilitate an easier visualisation of the key characteristics of the samples. There are

different options and the User can choose which figures to acquire from a menu that repeats until the

P1

339 cm-1

P2

290 cm-1

P3

370 cm-1

P4

378 cm-1

P5

305 cm-1

ZnS

696 cm-1

SnS

190 cm-1

SnS2

312 cm-1

Cu3SnS4

320 cm-1

CuS

475 cm-1

Lower Boundry 332 281 360 372 297 688 183 308 313 463

Upper Boundry 344 296 371 382 312 707 197 318 328 485

Main Peaks Secondary Phases

ROI

21

User terminates the loop and the program. The subsections that follow are descriptions of the options

provided through the figure menu.

3.1.4.1 EDS Experimental Data cf. EDS Fitted Values

To analyse the accuracy of the EDS fit, a plot is provided that compares empirical values and fitted

values. The values from the EDS data file are used to calculate the ratios used to present compositional

information. These values are then plotted together with the values of compositional ratios acquired

from the fit-functions at the coordinates for the EDS and Raman measurements respectively. Figure

19 shows the three different sets of values and how well they coincide. With greater similarity

between the experimental and fitted values a higher accuracy of the EDS fitting process can be

confirmed. In the particular case of figure 17, CZTS sample 12798A is shown with polynomial fits

with four degrees for both X and Y for copper, three degrees for both X and Y for zinc, and three

degrees for both X and Y for tin.

Figure 19: The accuracy of the fit created by the EDS analysis is shown in the figure by plotting the original empirical

values (blue) with the fitted values at the coordinates of the Raman (orange) and EDS (yellow) measurements. The X axis

is the ratio between copper and tin weight percentages. The Y axis is the ratio between zinc and the sum of copper and tin

weight percentages. The title of the figure specifies the polynomial degrees for the different fits of the elements. The red

oval highlights the cluster of experimental points at circa Cu/Sn = 2.

If no Raman analysis has been performed the plot shown in figure 19 would not compute, so the

option of comparing only the experimental and fitted EDS composition data is provided. This means

that the fitting accuracy can still be analysed. Having the values corresponding to the Raman

coordinates gives an idea of the relation between the measurement coordinates.

3.1.4.2 PL Intensity vs. Coordinates or Composition

The PL intensity is presented graphically as a simple plot with intensity versus X and Y coordinates

as shown in figure 20 and is simply used to find which areas contain the highest or lowest intensity.

As explained earlier the actual value of the intensity is not itself important but the contrasts shown

between the different areas are of interest. To make these types of plots useful it is vital to keep careful

documentation of the sample coordinates in relation to where the different elements were applied to

22

the substrate. In other words, the User of the program would be able to find the relationship between

peaks and element concentration by keeping in mind how the sample was created. An example of this

is knowing in which areas different elements were applied when making the sample.

Figure 20: The PL intensity for sample 12798A plotted versus the samples coordinates. The X and Y axes are in millimetres

while the colour scale represents intensity which should be used comparatively within the sample and not quantitative.

A simple way of finding the relation between intensity and compositional make-up of the sample is

to plot intensity as a function of compositional ratios instead of the actual coordinates. How

compositional axes work will be explained in the next section.

3.1.4.3 Raman Peak Intensity vs. Coordinates or Composition

Raman as opposed to PL, is dependent on which wavelength the laser has and therefore, is not as

straight forward as PL. When the 785 nm laser is used for CZTS, all the main peaks are of interest

and by extension the secondary order parameters, in other words, the ratios of peak intensities. The

first secondary order parameter is Q1, which is the intensity of main peak two divided by the intensity

of main peak five. Q2 is the second secondary order parameter, and it is calculated by dividing the

intensity of main peak one by the sum of the intensities of main peaks three and four. Also, the

secondary phase at the Raman shift of 320 cm-1, corresponding to Cu3SnS4, is shown when using a

laser with 785 nm wavelength. The program will only plot figures that are relevant according to the

wavelength of the laser used for the current data. The plots themselves will have the compositional

values on the X and Y axes while the third dimension is used for the intensity of the Raman at the

region of interest for the peak or phase in question. For CZTS the composition ratios are Zn divided

by the sum of Cu and Sn, and the ratio between Cu and Sn. As figure 21 displays, both the X and Y

axes are showing ratios of elemental atomic percentages. The areas where main peak one has the

largest influence can be seen in figure 21 as the brighter sections.

23

Figure 21: Raman intensity plotted versus compositional ratios for sample 12798A. The intensity is described by the colour

scale to the right and should be used as a comparative tool and not quantitative. The X axis is the proportion between Cu

and Sn weight percentage. The Y axis is the proportion between Zn weight percentage and the sum of Cu and Sn weight

percentages. The surface represents the samples compositional shape and is dependent on the weight percentage

concentration.

The option of plotting intensity versus the actual X and Y coordinates is also available, in the same

way the PL can be plotted versus coordinates. This is helpful in case a certain region is particularly

interesting. The User can, at a later stage, choose to look at the specific Raman spectra at that region

by giving the program the coordinates in X and Y. To get the X and Y coordinates the User needs to

look at the figures where intensity is plotted versus X and Y coordinates, instead of versus

compositional ratios.

3.1.4.4 Raman Spectrum at Specific Measurement Points

The evaluation of the Raman intensity plots can point to interest in a certain point or area of a sample,

as mentioned above. If the User is interested in the Raman spectrum pertaining to this point or area

this last option of plots can be used. By acquiring the coordinates with the Data Cursor from the

Raman Intensity Peaks vs. Coordinates plots the User can plot a graph of the Raman spectrum at these

points. The program askes for the number of points and then the coordinates for these points. Figure

22 is an example of how the result could look. In this case there are six measurement points, specified

in the legend, plotted for CZTS sample 12798A with a laser of wavelength 785 nm used for the

measurements.

24

Figure 22: The Raman intensity is plotted versus the wavenumber [cm-1] for six different measurement points. This graph

is an example of the result from the last section of the database tool program. The coordinates are shown in the legend in

the top right corner. The sample is 12798A and the wavelength of the laser used for the Raman experiment is 785 nm.

This finalizes the description of the different types of figures and graphs available. Appendix 1 will

be a presentation of all figures and graphs that can be provided by the program for a given sample.

25

4. Discussion The experience with the code when used in real cases have shown a satisfactory generation of the

graphs and figures required, with acceptable running time. The time it takes to perform different

actions like calculating the Raman intensity integrations or plotting the secondary order parameters

takes no more than a few seconds. The following sections discuss the EDS fitting and ROIs definition

to provide acceptable conditions for analysis by the User.

4.1 EDS fitting During the development of the fitting functionality it was observed that, in the case of the CZTS

samples, the surface fit for copper was significantly more demanding than for zinc and tin. The reason

for the need of a fourth-order, instead of third-order can be found in figure 26a (Appendix 1) and

figure 19. Both figures show the scatter of compositional ratios. The X axis shows the ratio of copper

with respect to tin. That means that the right side of the figures roughly show an elevated level of

copper. When examining these figures (26a and 19), the cluster of blue points at Cu/Sn = 2 is visible

and highlighted by the red oval in figure 19. This also gives reason to believe that the rearrangement

due to annealing is as expected. To the right of this cluster, the blue points become much more

scattered. This means that for these higher values of copper the fit has a more difficult time

encompassing the experimental data. This affects the surface fits directly. Figure 16b shows how the

fit is not following the measurement point values at X circa 40 mm and Y circa 10 mm. This is where

there is a high level of copper and the fit is only of order three in both X and Y. Figure 16a shows the

improvement when increasing to an order of four in both X and Y for that surface fit. The reason for

the difference in orders for the elements is because of the cluster of points and subsequent thinning

of points towards the copper intense section.

It can be concluded that in the process of developing an EDS fitting, when adding a new sample type,

the User should take careful stock of the compositional data to identify a potential problematic

element. Of course, there might be multiple problematic elements but either way, this needs to be

accounted for when choosing the orders of the polynomial fits. Specifically, for CZTS, there might

be a step in the production process of the samples that leads to the behaviour that copper has exhibited

here. Having poor surface fits could lead to the User getting bad compositional information when

looking at areas of interest on the samples. Worst case scenario would be that false conclusions are

drawn from the analysis due to flawed information caused by imprecise fits.

4.2 Region of Interest Determination Deciding which regions of interest to utilize for the different phases was expected to have an impact

on the calculated intensities, as opposed to the predefined intervals as explained in the results section.

The secondary order parameters Q1 and Q2 were expected to show some sort of difference in this

respect. This is because the main peaks used for the calculations of the parameters are close to each

other. That means that the ratios would be directly affected if a peak was used multiple times or not,

due to overlapping intervals for the ROI. Even having made sure the main peaks’ ROIs don’t overlap

there is still the issue of Cu3SnS4 which overlaps with main peak 5. This can have a significant impact

on Q1 due to that ratio being the intensity of main peak 2 divided by the intensity of main peak 5,

which is difficult to avoid.

Having an appropriate ROI can in some cases provide significant changes to the secondary order

parameters. Figure 23a and 23b show Q1 and Q2 plotted versus the compositional ratios when ROI

was seven wavenumbers in each direction of each main peak. Figures 24a and 24b are the same, with

the difference being that the ROI was the one decided in the results section. The observable difference

26

between figures 23a and 24a in terms of contrast is quite small. The same can be said for figures 23b

and 24b.

Figure 23 a & b: The secondary order parameters Q1 (a, left image) and Q2 (b, right image) plotted versus the compositional

ratios for CZTS sample 12798A. The X axis is the ratio of Cu and Sn. The Y axis is the ratio of Zn and the sum of Cu and

Sn. Q1 and Q2 were calculated with a set ROI of seven wavenumbers in each direction for all the main peaks.

Figure 24 a & b: The secondary order parameters Q1 (a, left image) and Q2 (b, right image) plotted versus the compositional

ratios for CZTS sample 12798A. The X axis is the ratio of Cu and Sn. The Y axis is the ratio of Zn and the sum of Cu and

Sn. Q1 and Q2 were calculated with a ROI that was described in table 1 for all the main peaks.

It is expected that the two secondary order parameters should show similar intensity distributions if

not exactly the same. This is obviously not the case for the figures above (23 and 24). The only

similarity seen is the pattern. For all these figures the sample can be roughly divided into three

sections. The first being where Cu/Sn is larger than two. The second where Zn/(Cu+Sn) is larger than

0.25 and Cu/Sn less than two. The last section is the left-over part. Even though the sections coincide

for Q1 and Q2, they do not follow the same proportional values. Where the highest ratio can be found

for Q1 the second highest is found for Q2. The reasons for this discrepancy could be due to

interference of peaks and phases. Particularly the interference of Cu3SnS4 on main peak 5, mentioned

earlier. This becomes a more difficult problem to solve if the identification of peaks is not done with

enough precision. Since the ROI was decided visually it is inherently prone to varying accuracy. Also,

main peak 1 can be affected by Cu3SnS4 due to the intervals being close.

If Cu3SnS4 is in fact affecting main peak 1 then Q2 will be affected as well. This can be seen when

comparing figures 28c and 30f (Appendix 1). The hotspot is found in the same place. That means that

the peak for Cu3SnS4 is probably affecting the integration of main peak 1. When looking at figure 24a

27

the place where Cu3SnS4 had its hotspot (in figure 30f) there should be a dark spot due to the ratio of

Q1 if Cu3SnS4 is having a significant impact on the ratio. This is not observed and therefore, the

conclusion is that Q2 is more affected by interference of the Cu3SnS4 peak, than Q1. This is why, at

least for this sample case, Q1 is more reliable than Q2 and it should be the one used when evaluating

a sample. When having another sample type i.e. not CZTS there might be similar problems with peaks

interfering. This means that even when the ROIs are chosen with care, so the intervals do not overlap

there might still be secondary order parameters that become useless due to an inability to separate the

peaks. This is where a peak fitting could be advantageous if it could make clearer differences between

the peaks than the ROIs do. When using CZTS samples, always check Q1 and Q2 versus Cu3SnS4 to

identify if any significant interference exists.

28

5. Conclusions The goal of the project has been achieved with regards to creating a code that can function as a

database and handle different types of data inputs for combinatorial synthesis analysis. Also, there is

room for expanding the program and code as well as an instruction file. A key strength of the CSDT

is that it is not limited to absorber layers. Other samples with other purposes that are analysed with

EDS, Raman or PL can be analysed with the program. Another key strength is the relative speed of

the program. Another strength lies in the fact that many of the choices made when running the

program are done with self-explanatory menus which makes it less error prone. A weakness of the

program, maybe the most important one, is that when adding a new sample type to database and code,

extra care is required. The instruction file is written to try and prevent these mistakes, but human error

can be difficult to eliminate.

5.1 Future Development It is possible to future develop the program code. It is in practice inevitable considering that adding a

new type of sample (i.e. a new material type) requires updates to the core code. The way to add a new

sample is described in the instruction document (see Appendix 2). The document describes the new

files that need to be created and the areas of the existing code that need to be updated. These updates

include choosing the order of the polynomial fit for the EDS data, adding secondary order parameter

calculations if necessary and adding new labels for the plots and figures.

When plotting intensities, a logarithmic scale could be used because it is the proportions of intensity

that are of interest and not the actual values. This would require the code involving plotting to be

changed from linear to logarithmic of course.

The EDS analysis step needs the User to put in eight values that correspond to the number of

measurement points and their placement on the sample. This could be considered tedious or

inefficient. Therefore, there could be a need for saving these particular values in some way that they

are tied to their EDS data set. This could then be used automatically by the program and save time.

One important aspect of this is that there might be multiple EDS data sets for one specific sample. In

other words, the correct values need to be paired to the correct data set when performing the EDS

analysis.

The program uses many Try and Catch blocks to allow the program to continue even if a certain

execution failed. These types of blocks have not been used on the data import sections. This is a step

that can malfunction and consequentially stop the entire program. One example of malfunction would

be if the data set is incomplete and the number of measurement points do not make a rectangular area.

One last suggestion for future development of the Combinatorial Synthesis Database Tool is the

extraction of data from the figures and graphs provided. Saving a figure in MATLAB is a trivial

matter, but extracting the actual data is more complicated. It could be advantageous to somehow

automate the process of saving the data.

5.2 New Insights Lessons learned are that when writing a code there may be multiple valid options and that progress is

seldom straight-forward. In multiple occasions a significant amount of time was spent on one method

for solving a problem when later it was proved a bad choice and another, better method was found.

This was naturally very frustrating but also educational. The progress could appear to be stopped at

certain points during the project, only for massive leaps to be made within a fraction of the time spent

29

stuck. Also, writing about a product in the form of a report has been challenging and developing. It

was much more difficult and time consuming than expected.

30

References Abou-Ras, D., Kirchartz, T. & Rau, U. (2011). Advanced Characterization Techniques for Thin-film

Solar Cells. Weinheim, Germany

Adolfsson, J. (2018). Characterization of combinatorial Cu2ZnSnS4 thin films. Uppsala University.

Solid-State Electronics (Master Thesis – Energy Systems Engineering Program)

Davydova, A., Rudisch, Katharina. & Scragg, J. (2011). Flexibility of the single phase region in

Cu2ZnSnS4 thin-films: theory and experiment. Chemistry of Materials. vol 23, 4625-4633

Koinuma, Hideomi & Takeuchi, Ichiro. (2004). Combinatorial Solid-State Chemistry of Inorganic

Materials. Nature Materials, vol 3, 429-438

Schmalensee, R., V. Bulovic, R. Armstrong, C. Batlle, P. Brown, J. Deutch, H. Jacoby, R. Jaffe, J.

Jean, R. Miller, F. O'Sullivan, J. Parsons, J.I. Pérez-Arriaga, N. Seifkar, R. Stoner and C. Vergara

(2015). The Future of Solar Energy: An Interdisciplinary MIT Study. Massachusetts Institute of

Technology, MIT Energy Initiative, May (http://mitei.mit.edu/futureofsolar)

31

Appendix 1 Each figure and graph available for a sample will be presented in this appendix. The data used corresponds to CZTS sample 12798A.

Figure 25 a, b & c: Cu, Sn and Zn surface fits for the EDS data plotted versus X and Y coordinates. The surface fits where done with polynomial expressions of order four in both X

and Y for Cu (left figure, a), order three in both X and Y for Sn (middle figure, b) and Zn (right figure, c). CZTS sample 12798A.

Figure 26 a & b: The composition ratios for the experimental EDS data presented together with composition ratios for EDS values acquired from the fit-functions. The values plotted

from the fit-functions are at the coordinates for Raman and EDS measurements. The left figure (a) only has the values at the EDS measurement points. The right figure (b) also has

the values at the Raman coordinates. CZTS sample 12798A.

32

Figure 27 a & b: PL intensity plotted versus composition ratios in the left (a) and versus coordinates in the right (b). CZTS sample 12798A.

Figure 28 a, b, c & d: Secondary order parameters Q1 and Q2 plotted versus composition (a & c) and versus coordinates (b & d). CZTS sample 12798A.

33

Figure 29 a, b, c, d, e & f: The left figures top (a) and bottom (d) are the Raman intensity for main peak one versus coordinates and composition respectively. The middle two (b &

e) are the same as a and d but for main peak two. The two to the right (c and f) are the same but for main peak three. CZTS sample 12798A.

34

Figure 30 a, b, c, d, e & f: The left figures top (a) and bottom (d) are the Raman intensity for main peak four versus coordinates and composition respectively. The middle two (b &

e) are the same as a and d but for main peak five. The two to the right (c and f) are the same but for the secondary phase Cu3SnS4. CZTS sample 12798A.

35

Figure 31 a, b, c, d, e & f: The left figures top (a) and bottom (d) are the Raman intensity for secondary phase ZnS versus coordinates and composition respectively. The middle two

(b & e) are the same as a and d but for secondary phase CuS when using a laser of 325 nm wavelength when performing the Raman measurements. The two to the right (c and f) are

the same as b and e but the laser used for the measurements hade wavelength 532 nm. CZTS sample 12798A.

36

Figure 32 a, b, c & d: The two left images (a & c) are the Raman intensity for secondary peak SnS versus coordinates and composition respectively. The two right images (b & d)

are the Raman intensity for secondary peak SnS2 versus coordinates and composition respectively. The reason a and c are zero is because the Raman data for the CZTS sample

12798A is cut of at 200 cm-1 which means that SnS that has its peak at 190 cm-1 has no data.

37

Appendix 2

Instruction File for Database Tool Database Tool for Combining, Storing and Analysing Material Data from Combinatorial Synthesis

By: Luciano Quaglia Casal

The database tool is made for handling data acquired from Raman Spectrometry, Energy-Dispersive

X-Ray (EDS) and Photoluminescence (PL) mapping. The User will be able to analyse these

measurement data both individually and collectively. Different plots are available with the aim of

facilitating visual inspection of the data. The data is collected from measurements done on different

combinatorial synthesis samples of absorber layers investigated for thin film solar cells.

The main code for the program is dbtool.m which is run by the User and implements a combination

of command line prompts and pop-up menus for User interaction. There are functions file that contain

specific sample type information, as well as lists of sample information with regards to the database.

The information in the function files is used in the analysis, while the information in the lists is for

the User to know which analysis options are available.

Adding New Sample Type Adding a new sample type to the database requires certain initial additions and updates to the main

code. Also, a new function file and a couple of other files need to be created, while a few of the other

existing files need to be amended. First, a new sample folder needs to be created in the main database

folder.

Files that need updating:

• dbtool.m

• ImpEDS.m

• EDSFit.m

• RAMANcalc.m

Always when updating a file there is a possibility that a new variable will be created and that means

that the clear commands that delete temporary variables need to be extended.

dbtool.m

Section 1 needs the menu sm and variable LoS to be updated with the new sample name option while

clear needs extending.

Section 2 needs the variable st to be updated with the new sample type name. All the menus called

dam need to be updated with the new sample type name as well.

Section 3 needs the variable st to be updated with the new sample type name. All the menus called

dam need to be updated with the new sample type name as well. All the if-loops that need the dam

need to be extended with elseif statements that take into account the new sample types. These elseif

statements need to follow the same form as the other options already specified. This means that the

outputs need to have the same variable names.

Section 4 has many if-loops that need to be extended with regards to the variable dam. The first two

are the most difficult to make work. The element variables, labels, titles and legends need to be

38

changed. The rest of the if-loops are mostly just updating the x- and y-labels. Plot menu 5 and 6 have

a very important if-loop that needs to be extended with the new sample. Variables nopMP and nopOP

need to be decided. They are the number of plots for Main Peaks and Other Phases respectively. This

is with regards to the different wavelengths of the lasers used for Raman measurements. This needs

a new if-loop separating the different wavelength options. Also, if secondary order parameters need

to be plotted they need to be specified here.

ImpEDS.m

There is one if-loop that is dependent on the variable dam that needs to be amended. The already

existing routines can be copied but the number of elements needs to be accounted for. Also, the Molar

Weights need to be specified to match the new elements. Changing the number of elements means

that the conversion from weight percentage to atomic percentage needs to be corrected.

EDSFit.m

This file can be complicated to update! The if-loop separates the different sample types and therefore,

elseif statements need to be added with respect to the variable dam. It is possible to copy the already

existing option and then change the element names and variable names. Most important is that the

result of the fitting process is still called fitresult and is in the same form. Matrixes need to be initiated

for the different elements according to the formatted EDS files. Theses matrixes are then filled with

the data from the formatted EDS file. Then the fitting for each element is performed and it is here that

the polynomial degrees are decided. Don’t forget to update the labels, titles and legends of the new

plots!

RAMANcalc.m

The Raman analysis calculation file has one if-loop that needs to be updated when adding a new

sample. It corresponds to the option of having secondary order parameters calculated. This goes hand

in hand with the code added to plot section (4) of the dbtool.m file. Make sure the variable names

coincide!

Files that need creating:

• Sample function file (ex. of existing function file: CZTSfun.m)

• Sample list importing file (ex. of existing function file: ImpSampInfoCZTS.m)

• Sample list in the form of a .txt file (ex. of existing function file: SampleInfoCZTS.txt)

The files that need creating can all be based on an existing file of the same sort. So, the sample list

just needs the name to have the right sample type and have an example row to illustrate the form each

row should take. This sample row also needs to be added to the total sample list (SampleInfo.txt) in

the main database folder. The import file can be copied with the new sample type taking the place of

the old. The filename variable needs the new sample folder name and the output variable creation

needs the new name. The function file is less generic and therefore needs more changes. This can still

be done by looking at an existing function file and working from there. This would include the main

peaks, other phases and the composition ratios for the plots.

39

Summary of Program The program starts by asking if the User would like to read this instruction document. This is

answered with either y for yes or n for no on the command line. Then the program displays a list of

the sections within the code. The first section is Sample Information, which gives the User the option

of plotting specific sample information lists and/or the main sample information list. This is done

through a menu pop-up.

Next is section two which is the Data Management section and runs through the database options.

These alternatives are adding a new sample folder, adding data to a sample folder and deleting whole

samples or specific files. At the same time information is added or changed in the relevant sample

lists.

Section three is the Analysis section and gives the User the option of analysing EDS, Raman and/or

PL data for a sample. When running through the EDS analysis, the data file used is formatted (if it

previously wasn’t) and then fitted with polynomial fits for the elements involved in the sample type.

The User is prompted to specify with “y” or “n” if the file is already formatted and then choosing the

formatted version for analysis. This is done in the command window. The results are surface-fit plots

and polynomial expressions that will be used to specify chemical composition at other measurement

points.

For Raman the User needs to choose the file to analyse as well as specifying the wavelength of the

laser used to acquire the data. The results being integration over peaks of interest. Analysing PL

requires the User to specify the number of files that make up a specific data set. This is because the

measurements can be made in multiple steps. The User is prompted on the command line to specify

the number of files and which files these are. The result of analysing PL is the integration of the whole

measured spectrum for each measurement point.

Figures and Graphs is the last section of the program and this is where the User chooses which types

of results are wanted. There are 6 types of plots available as listed below.

1. EDS experimental data cf. fits at Raman and at EDS measurement coordinates

2. EDS experimental data cf. fits at EDS measurement coordinates

3. PL Intensity vs. Composition

4. PL Intensity vs. X & Y Coordinates

5. Raman Intensity Peaks vs. Composition

6. Raman Intensity Peaks vs. X & Y Coordinates

7. Raman spectrum at specific measurement points

40

Appendix 3

dbtool.m

41

42

43

44

45

46

47

48

49

50

51

52

53

54

55

ImpEDS.m

56

ImpPL.m

57

EDSFit.m

58

59

60

RAMANcalc.m

61

CZTSfun.m

62

ImpSampInfoTOTAL.m

63

ImpSampInfoCZTS.m